Unified sending method and system for multiple types of mobile messages

Abstract

The invention discloses a unified sending method and system for multiple types of mobile messages, and relates to the technical field of artificial intelligence, and the method comprises the steps: responding to a message sending request, and obtaining the multiple types of mobile messages; inputting the multi-type mobile messages into a pre-trained message intelligent pushing iteration model, and outputting a final pushing combination; and packaging a response message of the message sending request based on the final push combination. The problem that an optimal push strategy cannot be generated for multiple types of mobile messages in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a unified sending method and system for multiple types of mobile messages. BACKGROUND

[0002] With the popularity of mobile Internet, multiple types of mobile messages such as short messages, APP push, instant messaging messages have become the core carrier of enterprise and user interaction. In order to integrate multi-channel resources, a unified message platform emerges as the times require, but the existing technology still has significant defects: first, the formats of multiple types of mobile messages are quite different (such as text, picture-text, link, etc.), and there is a lack of standardized processing mechanism, resulting in low message parsing efficiency and poor cross-channel adaptability; second, the push strategy is mostly fixed configuration, which only matches users based on a single dimension, and cannot realize accurate touch based on user portrait and historical response data, resulting in low message touch rate and conversion rate; third, there is a lack of dynamic optimization mechanism, and the push effect cannot be iteratively adjusted according to actual feedback, which may cause user fatigue and response decline after long-term use.

[0003] In addition, some schemes do not design differentiated push logic for different types of messages, and lack channel priority sorting and small flow test verification links, which are difficult to meet the enterprise's large-scale and high-precision message sending needs. Therefore, there is an urgent need for a method capable of generating an optimal push strategy for multiple types of mobile messages. SUMMARY

[0004] Therefore, the present application provides a unified sending method and system for multiple types of mobile messages, which can generate an optimal push strategy for multiple types of mobile messages.

[0005] To achieve the above purpose, the present application provides the following technical scheme: A unified sending method for multiple types of mobile messages, comprising: In response to a message sending request, obtaining multiple types of mobile messages; Inputting the multiple types of mobile messages into a pre-trained message intelligent push iterative model, the message intelligent push iterative model comprising a receiving layer, a processing layer, an extraction layer, a label generation layer, a matching layer, an inventory generation layer, a text generation layer, a test layer, a feature extraction layer, an iterative optimization layer and an output layer; Receiving multiple types of mobile messages through the receiving layer; Performing data processing on the multiple types of mobile messages through the processing layer to obtain standardized messages in a unified format; Extracting core information of the standardized messages through the extraction layer; Generating a to-be-pushed message label based on the core information through the label generation layer; retrieve a target user group matched with the to-be-pushed message label from a user portrait database through a matching layer, and extract available target accounts of the target user group on each platform; perform multi-dimensional quantitative scoring on each available target account through a list generation layer to obtain an available target account routing sorting list of each target user; generate a corresponding push text for each available target account in the available target account routing sorting list through a text generation layer; perform a small-flow push test through a test layer to obtain an optimal push combination; perform feature extraction on the optimal push combination through a feature extraction layer to obtain optimal push features; optimize a push strategy of a message intelligent push iteration model based on the optimal push features through an iterative optimization layer until a push effect meets a preset termination condition; output a final push combination through the output layer; encapsulate a response message of the message sending request based on the final push combination.

[0006] Optionally, the core information of the standardized message is extracted through the extraction layer, including: perform word segmentation processing on the standardized message to obtain a plurality of keywords; calculate a semantic matching score and a keyword matching score of the keywords; calculate a comprehensive confidence of the keywords based on the semantic matching score and the keyword matching score; when the comprehensive confidence is greater than a preset threshold, determine the keywords as core information.

[0007] Optionally, the to-be-pushed message label is generated based on the core information through the label generation layer, including: generate a fixed label for the core information based on a preset classification rule; generate a dynamic label for the core information based on user portrait data and a collaborative filtering algorithm; combine the fixed label and the dynamic label to generate a to-be-pushed message label.

[0008] Optionally, the multi-dimensional quantitative scoring on each available target account is performed through the list generation layer to obtain an available target account routing sorting list of each target user, including: calculate an account validity score, a scene adaptability score, a user preference score, a historical response rate score, and a compliance status score of each available target account; calculate a comprehensive score of each available target account based on the account validity score, the scene adaptability score, the user preference score, the historical response rate score, and the compliance status score. For all available target numbers of the same target user, sort them according to the comprehensive score and generate a routing sort list of available target numbers for that user.

[0009] Optionally, generating corresponding push text for each available target number in the available target number routing sort list through the text generation layer includes: Based on the platform characteristics of the available target number and the tag of the message to be pushed, an appropriate template is selected from the preset template library, and the core information is automatically filled into the appropriate template according to the preset field positions; Based on the account adaptation characteristics of different available target numbers for the same user and the interaction scenarios of the platform, differentiated push text is generated; Associate available target numbers with differentiated push text.

[0010] Optionally, the step of conducting small-volume push tests through the testing layer to obtain the optimal push combination includes: A stratified random sampling algorithm was used to select users from different user profile dimensions as test users. A random allocation mechanism with a single account and a single text version is adopted to randomly assign available target numbers and push texts to the test users. Each test user receives only one push text corresponding to one available target number. After starting small-volume push notifications, the comprehensive score for each push combination is calculated using a multi-indicator weighted scoring formula; If the overall score is greater than the preset value, the push combination is determined to be the optimal push combination and is output directly. If the overall score is less than a preset value, the text generation layer is triggered to regenerate differentiated push text until the optimal push combination with an overall score greater than the preset value is obtained.

[0011] Optionally, the step of extracting features from the optimal push combination through a feature extraction layer to obtain the optimal push features includes: Extract the semantic gene features, text structure features, information emphasis features, and account adaptation features of each optimal push combination, and calculate the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is generated based on the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is output to the iterative optimization layer as an adjustment to the push strategy.

[0012] A unified system for sending multiple types of mobile messages, comprising: The acquisition module is used to retrieve various types of mobile messages in response to message sending requests; The push combination determination module is used to input the multi-type mobile messages into a pre-trained intelligent message push iterative model. The intelligent message push iterative model includes a receiving layer, a processing layer, an extraction layer, a tag generation layer, a matching layer, a list generation layer, a text generation layer, a testing layer, a feature extraction layer, an iterative optimization layer, and an output layer. Receive multiple types of mobile messages through the receiving layer; The processing layer processes the various types of mobile messages to obtain standardized messages in a unified format. The core information of the standardized message is extracted through the extraction layer; The tag generation layer generates tags for messages to be pushed based on the core information; The matching layer retrieves the target user group that matches the tag of the message to be pushed from the user profile database, and extracts the available target accounts of the target user group on each platform. The list generation layer performs multi-dimensional quantitative scoring on each available target number to obtain a sorted list of available target numbers for each target user. The text generation layer generates corresponding push text for each available target number in the sorted list of available target numbers; The optimal push combination was determined by conducting small-scale push tests through the test layer. The optimal push features are obtained by extracting features from the optimal push combination through a feature extraction layer. The push strategy of the intelligent push iterative model is optimized by the iterative optimization layer based on the optimal push features until the push effect meets the preset termination condition. The final push combination is output through the output layer; The response module is used to encapsulate the final push combination into a response message for the message sending request.

[0013] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method described herein.

[0014] A non-transitory computer-readable storage medium having a computer program stored thereon, the computer program implementing the steps of the method when executed by a processor.

[0015] The present invention has the following advantages: This invention presents a unified method for sending multiple types of mobile messages. Through a multi-level collaborative design of an intelligent message push iterative model, it achieves standardized processing and accurate extraction of core information for various types of mobile messages, ensuring targeted push notifications. It combines user profiles with target accounts, generates a routing ranking list through multi-dimensional quantitative scoring, and pairs it with personalized push text to improve message adaptability. Optimal push combinations are selected through small-scale testing, and the strategy is dynamically adjusted through feature extraction and iterative optimization layers to continuously optimize push performance. This effectively solves the problems of rigid push strategies and poor adaptability in existing technologies, significantly improving message reach and user interaction willingness, reducing operating costs, meeting the needs of large-scale, high-precision multi-type mobile message sending, and improving user experience and platform service quality. Attached Figure Description

[0016] For illustrative and not limiting purposes, the present invention will now be described in conjunction with embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the unified sending method for multiple types of mobile messages in an embodiment of the present invention. Figure 2 This is a schematic diagram of the main components of a unified sending system for multiple types of mobile messages in an embodiment of the present invention; Figure 3 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] It should be noted that, where there is no conflict, the embodiments and features of the present invention can be combined with each other. The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating the unified sending method for multiple types of mobile messages in an embodiment of the present invention, such as... Figure 1 As shown, the unified sending method for multiple types of mobile messages provided in this embodiment of the invention includes the following steps S101 to S103.

[0021] S101, in response to a message sending request, retrieve multiple types of mobile messages.

[0022] Message sending requests can be initiated by the business system (such as a repayment reminder triggered by bill generation), triggered by user actions (such as a receipt notification triggered after a user applies for a service), or triggered by a scheduled task (such as a service notification pushed at a fixed time every day).

[0023] Through preset interfaces (such as API, database synchronization, message queue), various types of mobile messages are collected from multiple platforms such as APP, SMS platform, WeChat official account, and email system. These mobile message types include information such as text content, business identifier, sending timestamp, and associated user ID.

[0024] The system performs format validity and field integrity checks on the collected mobile messages of various types, filters out duplicate messages and invalid messages that are missing core fields (such as user ID and business type), and ensures the validity of data for subsequent processing.

[0025] S102 inputs multiple types of mobile messages into a pre-trained intelligent message push iterative model to output the final push combination.

[0026] The intelligent message push iterative model includes a receiving layer, a processing layer, an extraction layer, a tag generation layer, a matching layer, a list generation layer, a text generation layer, a testing layer, a feature extraction layer, an iterative optimization layer, and an output layer. Receive multiple types of mobile messages through the receiving layer; The processing layer processes the various types of mobile messages to obtain standardized messages in a unified format. The core information of the standardized message is extracted through the extraction layer, including: The standardized message is segmented to obtain several keywords; A hybrid word segmentation algorithm that integrates industry terminology dictionaries is adopted, specifically "jieba word segmentation core algorithm + custom business terminology dictionary". The custom business terminology dictionary contains exclusive terms for each business scenario (such as "bill installment" and "credit limit adjustment" in the financial scenario, and "order verification" and "logistics tracking" in the e-commerce scenario), ensuring that industry terms are not split. After tokenizing the standardized messages, filter out invalid words through a preset stop word library (including common stop words such as "de", "ji", "le" and business-irrelevant words such as "notification", "reminder" in the basic expressions), and retain entity words, action words, and business attribute words as candidate keywords; Perform词性标注(如名词、动词、形容词),优先保留名词类(如“账单”“账号”)和动词类(如“申请”“还款”)关键词,进一步提升筛选精准度。

[0027] Calculate the semantic matching score and keyword matching score of the keywords; In the formula, is the keyword matching score (the value range is 0-1, the higher the score, the higher the matching degree of the keyword with the business sub-library), is the number of hits of the candidate keyword in the corresponding business sub-library, is the preset weight of the i-th hit keyword (the value range is 0.1-1.0, the weight of the core business term ≥ 0.6, and the weight of the secondary attribute word ≤ 0.4), is the total number of keywords in the corresponding business sub-library (m ≥ 1, an integer), is the preset weight of the j-th keyword in the business sub-library (all The sum is the total weight of the sub-library to ensure the unity of the weight systems of different sub-libraries).

[0028] ; In the formula, is the semantic matching score (the value range is 0-1, the closer it is to 1, the higher the semantic fit degree), is the semantic vector of the candidate keyword (generated by encoding with an industry-specific NLP model, with a dimension of d, d ≥ 768), is the benchmark semantic vector of the corresponding business sub-library (generated by aggregating and encoding the typical core texts of the sub-library, with the same dimension as ; If the candidate keyword is a multi-word combination, is the average value of the semantic vectors of each word (weighted average according to the词性权重加权平均,名词权重0.6,动词权重0.4).

[0029] Calculate the comprehensive confidence of the keyword based on the semantic matching score and keyword matching score; ; In the formula, is the comprehensive confidence of the keyword (the value range is 0-1, used to determine whether it is core information), It should be noted that there are some incomplete or unclear expressions in the original text, such as "Perform词性标注(如名词、动词、形容词)" and "词性权重加权平均,名词权重0.6,动词权重0.4" which need to be further clarified in the original context for more accurate translation.This is the weighting coefficient for the semantic matching score (value range 0.6-0.8, default value 0.7, can be dynamically adjusted according to business scenarios).

[0030] When the overall confidence level is greater than a preset threshold, the keyword is determined to be core information; The tag generation layer generates tags for the messages to be pushed based on the core information, including: Based on preset classification rules, fixed labels are generated for the core information; Dynamic tags are generated for the core information based on user profile data and collaborative filtering algorithms; ; In the formula, The comprehensive weight of the t-th dynamic tag (the value ranges from 0 to 1, and the higher the weight, the higher the fit between the tag and the user / core information). The weighting coefficient for user profile matching accuracy (value range 0.5-0.7, default value 0.6, can be adjusted according to business scenarios). The matching degree between dynamic tags and user profiles (value range 0-1, calculated from multi-dimensional matching of user profiles). This is the collaborative filtering similarity for dynamic tags (values ​​range from 0 to 1, calculated based on the tag preferences of similar user groups).

[0031] The fixed and dynamic tags are combined to generate tags for the message to be pushed. The matching layer retrieves the target user group that matches the tag of the message to be pushed from the user profile database, and extracts the available target accounts of the target user group on each platform. The list generation layer performs multi-dimensional quantitative scoring on each available target number to obtain a sorted list of available target numbers for each target user, including: Calculate the account validity score, scenario suitability score, user preference score, historical response rate score, and compliance status score for each available target number; A comprehensive score for each available target number is calculated based on the account validity score, scenario adaptability score, user preference score, historical response rate score, and compliance status score. For all available target numbers of the same target user, sort them according to the comprehensive score and generate a routing sort list of available target numbers for that user. The text generation layer generates corresponding push text for each available target number in the available target number routing sort list, including: Based on the platform characteristics of the available target number and the tag of the message to be pushed, an appropriate template is selected from the preset template library, and the core information is automatically filled into the appropriate template according to the preset field positions; Based on the account adaptation characteristics of different available target numbers for the same user and the interaction scenarios of the platform, differentiated push text is generated; Associate available target numbers with differentiated push text; By conducting small-scale push tests through the test layer, the optimal push combination was obtained, including: A stratified random sampling algorithm was used to select users from different user profile dimensions as test users. A random allocation mechanism with a single account and a single text version is adopted to randomly assign available target numbers and push texts to the test users. Each test user receives only one push text corresponding to one available target number. After starting small-volume push notifications, the comprehensive score for each push combination is calculated using a multi-indicator weighted scoring formula; If the overall score is greater than the preset value, the push combination is determined to be the optimal push combination and is output directly. If the overall score is less than the preset value, the text generation layer is triggered to regenerate the differentiated push text until the optimal push combination with an overall score greater than the preset value is obtained; The optimal push combination is subjected to feature extraction through a feature extraction layer to obtain the optimal push features, including: Extract the semantic gene features, text structure features, information emphasis features, and account adaptation features of each optimal push combination, and calculate the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is generated based on the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is output to the iterative optimization layer as an adjustment to the push strategy. The push strategy of the intelligent push iterative model is optimized by the iterative optimization layer based on the optimal push features until the push effect meets the preset termination condition. The final push combination is output through the output layer.

[0032] S103, based on the final push combination, encapsulates into a response message for the message sending request.

[0033] Figure 2 This is a schematic diagram illustrating the main components of a unified sending system for multiple types of mobile messages according to an embodiment of the present invention. For example... Figure 2 As shown, the unified sending system 1 for multiple types of mobile messages provided in this embodiment of the invention includes an acquisition module 10, a push combination determination module 20, and a response module 30.

[0034] Module 10 is used to acquire multiple types of mobile messages in response to a message sending request; The push combination determination module 20 is used to input the multi-type mobile messages into a pre-trained intelligent message push iterative model to output the final push combination; The intelligent message push iterative model includes a receiving layer, a processing layer, an extraction layer, a tag generation layer, a matching layer, a list generation layer, a text generation layer, a testing layer, a feature extraction layer, an iterative optimization layer, and an output layer. Receive multiple types of mobile messages through the receiving layer; The processing layer processes the various types of mobile messages to obtain standardized messages in a unified format. The core information of the standardized message is extracted through the extraction layer; The tag generation layer generates tags for messages to be pushed based on the core information; The matching layer retrieves the target user group that matches the tag of the message to be pushed from the user profile database, and extracts the available target accounts of the target user group on each platform. The list generation layer performs multi-dimensional quantitative scoring on each available target number to obtain a sorted list of available target numbers for each target user. The text generation layer generates corresponding push text for each available target number in the sorted list of available target numbers; The optimal push combination was determined by conducting small-scale push tests through the test layer. The optimal push features are obtained by extracting features from the optimal push combination through a feature extraction layer. The push strategy of the intelligent push iterative model is optimized by the iterative optimization layer based on the optimal push features until the push effect meets the preset termination condition. The final push combination is output through the output layer; The response module 30 is used to encapsulate the final push combination into a response message for the message sending request.

[0035] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, the electronic device 40 includes: a processor 401, a memory 402, and a bus 403; The processor 401 and the memory 402 communicate with each other via the bus 403. The processor 401 is used to call program instructions in the memory 402 to execute the methods provided in the above-described method embodiments, and to execute the methods provided in the embodiments of the present invention.

[0036] This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions, which cause a computer to execute the method provided in this embodiment of the invention.

[0037] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various storage media capable of storing program code, such as ROM, RAM, magnetic disk, or optical disk.

[0038] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A unified method for sending multiple types of mobile messages, characterized in that, include: Responding to message sending requests, retrieve various types of mobile messages; The various types of mobile messages are input into a pre-trained intelligent message push iterative model to output the final push combination; The intelligent message push iterative model includes a receiving layer, a processing layer, an extraction layer, a tag generation layer, a matching layer, a list generation layer, a text generation layer, a testing layer, a feature extraction layer, an iterative optimization layer, and an output layer. Receive multiple types of mobile messages through the receiving layer; The processing layer processes the various types of mobile messages to obtain standardized messages in a unified format. The core information of the standardized message is extracted through the extraction layer; The tag generation layer generates tags for messages to be pushed based on the core information; The matching layer retrieves the target user group that matches the tag of the message to be pushed from the user profile database, and extracts the available target accounts of the target user group on each platform. The list generation layer performs multi-dimensional quantitative scoring on each available target number to obtain a sorted list of available target numbers for each target user. The text generation layer generates corresponding push text for each available target number in the sorted list of available target numbers; The optimal push combination was determined by conducting small-scale push tests through the test layer. The optimal push features are obtained by extracting features from the optimal push combination through a feature extraction layer. The push strategy of the intelligent push iterative model is optimized by the iterative optimization layer based on the optimal push features until the push effect meets the preset termination condition. The final push combination is output through the output layer; The final push combination is encapsulated into a response message for the message sending request.

2. The unified sending method for multiple types of mobile messages according to claim 1, characterized in that, The extraction of core information from the standardized message through the extraction layer includes: The standardized message is segmented to obtain several keywords; Calculate the semantic matching score and keyword matching score of the keywords; The overall confidence level of the keyword is calculated based on the semantic matching score and the keyword matching score; When the overall confidence level is greater than a preset threshold, the keyword is determined to be core information.

3. The unified sending method for multiple types of mobile messages according to claim 2, characterized in that, The step of generating push message tags based on the core information through the tag generation layer includes: Based on preset classification rules, fixed labels are generated for the core information; Dynamic tags are generated for the core information based on user profile data and collaborative filtering algorithms; The fixed and dynamic tags are combined to generate tags for the message to be pushed.

4. The unified sending method for multiple types of mobile messages according to claim 1, characterized in that, The process of generating a list to perform multi-dimensional quantitative scoring on each available target number, resulting in a sorted list of available target numbers for each target user, includes: Calculate the account validity score, scenario suitability score, user preference score, historical response rate score, and compliance status score for each available target number; A comprehensive score for each available target number is calculated based on the account validity score, scenario adaptability score, user preference score, historical response rate score, and compliance status score. For all available target numbers of the same target user, sort them according to the comprehensive score and generate a routing sort list of available target numbers for that user.

5. The unified sending method for multiple types of mobile messages according to claim 1, characterized in that, The step of generating corresponding push text for each available target number in the available target number routing sort list through the text generation layer includes: Based on the platform characteristics of the available target number and the tag of the message to be pushed, an appropriate template is selected from the preset template library, and the core information is automatically filled into the appropriate template according to the preset field positions; Based on the account adaptation characteristics of different available target numbers for the same user and the interaction scenarios of the platform, differentiated push text is generated; Associate available target numbers with differentiated push text.

6. The unified sending method for multiple types of mobile messages according to claim 1, characterized in that, The process of conducting small-volume push tests through the testing layer to obtain the optimal push combination includes: A stratified random sampling algorithm was used to select users from different user profile dimensions as test users. A random allocation mechanism with a single account and a single text version is adopted to randomly assign available target numbers and push texts to the test users. Each test user receives only one push text corresponding to one available target number. After starting small-volume push notifications, the comprehensive score for each push combination is calculated using a multi-indicator weighted scoring formula; If the overall score is greater than the preset value, the push combination is determined to be the optimal push combination and is output directly. If the overall score is less than a preset value, the text generation layer is triggered to regenerate differentiated push text until the optimal push combination with an overall score greater than the preset value is obtained.

7. The unified sending method for multiple types of mobile messages according to claim 1, characterized in that, The step of extracting features from the optimal push combination through a feature extraction layer to obtain the optimal push features includes: Extract the semantic gene features, text structure features, information emphasis features, and account adaptation features of each optimal push combination, and calculate the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is generated based on the semantic gene feature score, text structure feature score, information emphasis feature score, and account adaptation feature score. The optimal push feature vector is output to the iterative optimization layer as an adjustment to the push strategy.

8. A system for unified transmission of multiple types of mobile messages, characterized in that, include: The acquisition module is used to retrieve various types of mobile messages in response to message sending requests; The push combination determination module is used to input the multi-type mobile messages into a pre-trained intelligent message push iterative model to output the final push combination; The intelligent message push iterative model includes a receiving layer, a processing layer, an extraction layer, a tag generation layer, a matching layer, a list generation layer, a text generation layer, a testing layer, a feature extraction layer, an iterative optimization layer, and an output layer. Receive multiple types of mobile messages through the receiving layer; The processing layer processes the various types of mobile messages to obtain standardized messages in a unified format. The core information of the standardized message is extracted through the extraction layer; The tag generation layer generates tags for messages to be pushed based on the core information; The matching layer retrieves the target user group that matches the tag of the message to be pushed from the user profile database, and extracts the available target accounts of the target user group on each platform. The list generation layer performs multi-dimensional quantitative scoring on each available target number to obtain a sorted list of available target numbers for each target user. The text generation layer generates corresponding push text for each available target number in the sorted list of available target numbers; The optimal push combination was determined by conducting small-scale push tests through the test layer. The optimal push features are obtained by extracting features from the optimal push combination through a feature extraction layer. The push strategy of the intelligent push iterative model is optimized by the iterative optimization layer based on the optimal push features until the push effect meets the preset termination condition. The final push combination is output through the output layer; The response module is used to encapsulate the final push combination into a response message for the message sending request.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images